Free data transfer out to internet when moving out of AWS(aws.amazon.com) |
Free data transfer out to internet when moving out of AWS(aws.amazon.com) |
EU – actually, it must be free.
Big tech firms – free and open competition is integral to society, and we value freedom of choice over anything else, so of course data transfer is free for our valued customers.
> Where a data processing service is being used in parallel with another data processing service, the providers of data processing services may impose data egress charges, but only for the purpose of passing on egress costs incurred, without exceeding such costs.
Hopefully this article doesn't end up with exploitable loopholes. Bringing down AWS, GCP, and Azure egress costs to market rates could majorly help reduce cloud lock-in gradually without having to close your entire account.
[1]: https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=OJ:...
What exactly do they consider "costs"? Is it just the cost they pay an IP for data out or can it include maintenance, hardware, people, etc?
I am also curious if this will have any impact on the cost of services. Does AWS do any sort of thing where the higher egress costs offset the cost of services? Or ingress costs, or internal communication costs? Stuff that does cost AWS something to run in some form but could have been paid for by this.
It's a question I honestly don't know, so I am curious.
The second question is a bit trickier to answer. Amazon has a lot of fixed costs and ABC is one way to allocate those fixed costs (fixed costs = servers). You can also do something similar for the revenues of different AWS services. However, managers will tend to look at revenues and costs together because teasing apart the different revenue streams and their allocated cost structures doesn’t match reality. If you go too deep into that rabbit hole, you lose sight of the true nature of the business. Which is “we buy a shitload of servers up front and rent them to you virtually by the second.” To answer the question, the egress revenue likely dwarfs all the other revenue generating activities at AWS, making the other services look “bad” when comparing revenue to costs. On paper, egress makes up substantially all profit for the business. But this is by design- executives believe that this business model captures the most value so they’re not that interested in jacking up other service costs because overall profitability will be less. Is there subsidization? Yes, but Amazon believes they have the better business by charging users like this vs pricing everything “fairly.”
to a degree I have seen AWS plans work this way, where you choose from options A/B/C and buyer beware if you end up getting scalped.
In other words how much AWS is overcharging?
I guess they make every effort / creative bookkeeping trick to exaggerate their egress costs and it will take years until EU completes their investigation and fines them and more years until courts have had their final word.
AWS could plausibly claim _some_ additional expense on their premium network offering (not full rate, of course, but some) for the additional monitoring and adjustment they do to adjust to peering conditions but none of that is plausible for other companies peering with them who provide the same services. If their customer uses Fastly or Cloudflare, they don’t need any additional cost to toss packets over the wall and let the other service worry about routing. I’d imagine those companies are making sure that the EU regulators know exactly how much they need to charge for that.
And all their network costs includes all the routers, all the fibers, all the staff in the networking and infrastructure teams, and all the teams that support those teams.
Still a positive, but man, that marketing messaging. I give props to the team who wrote that.
Probably obeying the letter of the European Data Act, but obviously not great if true.
Does it really matter? As long as they are upfront about the price and there is no vendor lock in, what's the big deal? AWS is overpriced - everyone knows that, but i don't think its morally wrong to be overpriced as long as you aren't deceptive about it.
> We don’t require you to close your account or change your relationship with AWS in any way. You’re welcome to come back at any time. We will, of course, apply additional scrutiny if the same AWS account applies multiple times for free DTO.
> After your move away from AWS services, within the 60-day period, you must delete all remaining data and workloads from your AWS account, or you can close your AWS account.
Damn, that's a lot of tiny customers. 100 GB per month is less than what many use on their phone during a month. So basically, most people using AWS is definitely people who don't need it to scale.
Small companies download reports/analytics/files from X and upload them to Y, etc, big companies do it all inside.
This is what we would expect, right? Surely the huge majority of AWS users are small and wouldn't have the slightest interest in owning/managing their own hardware.
I think there's a good saying about Cloud and scaling, paraphrasing - "if you don't start on Cloud that's unwise and if you (over)stay that's insane". I guess as your solution reaches maturity and stability (no sudden swings in usage or complete changes in the architecture) then you should start thinking about some on-prem solution with maybe the hardware being leased along with maitenance.
The law mandates egress costs should either be free or not exceed the costs encountered by the cloud provider.
This is a particularly meaningless gesture in the direction of compliance from a provider dominant across so many categories of cloud services as AWS is -- the whole point of the "in parallel" requirement in the law is, I would assume, to prevent the mega-cloud providers (AWS, mainly, but also Azure and GCP) from being able to effectively lock customers to monopolize areas of cloud services where their specific offerings are weak by the egress fees of their dominant offerings which would be used alongside them. "Free only when you are transferring out to leave our platform" doesn't address this lock-in at all.
So briefly in this case AWS does something which seems superficially good, but the real reason is to appease a law that it actually doesn't properly comply with. (Is the, not my, claim.) i.e. a good thing done in bad faith.
https://www.cnbc.com/2024/01/11/google-vows-not-to-charge-da...
The managed services are expensive, but boy do they work well. Over the last couple years been running +100 each of managed elastic searches, RDS, and firehouse with minimal to no issues.
An interesting question: why does AWS charge so much for egress knowing that it could dissuade some currently-on-prem use cases? It definitely forces optimization, use of CDNs, compression, etc. Previously it likely reduced transfers to another provider, but as the parent alludes to, the use of managed services makes sticking to AWS much more likely. What else?
Edit: My search engine found December 2021:
https://aws.amazon.com/about-aws/whats-new/2021/11/aws-price...
At least for customers using a lot of egress the charging per address is probably cheaper. For small businesses with relatively little businesses but not so slim deployments with addresses hanging around on many little used or even forgotten resources it's the other way round.
Edit: I had already forgotten the 42% price raise hitting myself on my personal Lightsail instance in May.
https://www.sidn.nl/en/news-and-blogs/are-we-past-peak-ipv4
Amazon’s move seems to be motivated by pure greed and quasi-monopoly complacency.
A lot of multi-cloud architectures are unlikely to be designed because the egress fees are killing them. Imagine if we could host some of the infra in AWS, using S3, but keeping some machines on Hetzner, while calling other APIs hosted in azure… one day it will happen :’)
And some architectures are still prohibited - any replicated database across cloud vendors - imagine Cassandra hosted across AWS and Azure to protect against one cloud vendor outage.
Take EU261, which caused issues for smaller airlines, leading to less competition, increased fares. There are also claims it’s affecting safety [0]. I’m not against EU261 per se, but it’s healthy to be critical of large institutions like the EU.
People, I find, cheer about new EU legislation but then forget to look think about any consequences a few years down the road.
Where is the EU’s own critical evaluation and performance review of EU261, especially given it’s been heavily revised through case law over the years?
(FWIW, I live in Europe, I’m generally pro EU, but it’s not all roses)
[0] (obviously some bias given it’s the ERAA) https://www.eraa.org/sites/default/files/era_eu261_study_bro...
So, if it caused issues for smaller airlines and generally increase in prices, is that not a clear indication that the smaller airlines do not have sufficiently capability to handle it, and that prices was too low?
Otherwise, if they really cared about their customers, they'd just lower outgoing data transfer pricing at all times and not only when you're leaving.
Unless of course you went all in on IAM, Lambda, API gateway etc etc instead of just running software on Linux.
In which case you’ll need to stay.
No, you don’t.
Thank you, EU, on behalf of the rest of the world
Reality: the invisible hand is simply a fist made up of all members of a cartel.
Googles equivalent announcement was all flowery bullshit making it seem like a voluntary action
It shows they know they are raking people over the coals with intentional, preconceived illegal anticompetitive behavior written into their documentation.
Right from the playbook, they have a sweet marketing wrapup for when they have to walk back their preconceived illegal behavior to the minimum directed border.
Tech corps especially now do this as standard operating procedure (I mean Uber had entire divisions and armies of lawyers to bypass/obfuscate/delay imposition of local cab and labor laws). It's why you can't trust them with AI, personal information, AI, customer service, AI, following laws, AI, or anything.
I know Google in the heydey of the aughts wasn't as "good" as they tried to be, but at least the "do no evil" kind of kept them in check to some degree. But at some point they bought the corporation-formerly-known-as-DoubleClick, and I think in retrospect that was the beginning of the end of functional idealism in the internet.
I have to say, talking to AWS and GCP, the AWS marketing, onboarding, and support arm are far far far far far superior to google's. In fact they are so far apart that our management has blatantly discarded google as even an option for moving our infra over because of how terrible our interactions have been. I'm all for de-throning the king, but it ain't gonna be done by the court jester.
In any case, point is, AWS has good strategy where it matters most.
Speaking for and only for Amazon Redshift, as I have little knowledge of other AWS services, I hold AWS's blogs, messaging, Support communications, TAMs, the lot, as relentlessly positive and to my eye deliberately and knowingly obfuscating all weakness. I regard information from AWS regarding Redshift as safe to read when and only when you already know what's going on / the underlying truth. Otherwise you will be misled, and to your cost at AWS's benefit.
By the sounds of it, the messaging over this change in data policy is the same.
I've never read AWS blogs before but that was exactly my feeling after reading the linked article on "fair software principles" or something. You can tell that somebody has a hobby horse they're riding, and it's only because it's in their financial self-interest, but unless you already knew what the actual issue was before you read the blog, you're still going to be clueless afterwards!
You're strictly limiting yourself to companies that provide no internet-facing services?
With such a high price per IP there are a lot of organizations finally putting in the effort of cleaning up their space and selling off blocks. I know a company that was doing consolidated and clean up for an aggregation of freeing 200k worth of space. They were looking at adding $15-35M to their pockets.
IPv6 support is still not where it needs to be. Look at the clouds. There are many services that still are only IPv4 or partial ipv6 support. Just check back in 1-3 years and the price will have recovered.
It’s meant for situations where you intend to leave AWS for good, not spin up another EC2 instance when you need it, generate terabytes of data and then transfer it out for free.
I only have ~80 GB stored, and rather than being charged 44 cents per month, I'm being charged 88 cents every 2 months.
Plus, the egress fees were too damn high
https://accelerationeconomy.com/cloud-wars/amazon-ceo-andy-j...
Companies don’t like to have their bottom line affected.
I’m unconvinced. A lot of money is spent on third party services (think Snowflake, but there are a ton) where the third party service is offered in AWS, GCP, and Azure. There is no on-the-Internet option.
If egress fees go away, on-the-Internet becomes appealing, and there goes a lot of revenue. If fees for other services go up, on-the-Internet becomes even more appealing.
1) Starting charging legacy fees like the MySQL 5.7 fees 2) Start charging for hits, IO, etc 3) Tier out fees like with Aurora IO optimized to increase revenue 4) Push support levers 5) Stop reducing prices for older instance models
"Okay cool! We'll sign you up! Good luck!"
vs
"We have a migration strategy, and partners who can help you lift-and-ship or at least consult to get you a lay of the land. Try these companies and see if you're interested, and we'll help you along the way."
Google is -- thanks for reaching out, you can certainly sign up and figure it all out, here's a few books!
Amazon -- things are complex, and here's a few books to read, but I know you don't have a few years so here's some people who know how to do this, and we've ironed out how most companies succeed in migrating. Also he's a ton of incentive to move to amazon-exclusive services.
Amazon’s behaviour isn’t surprising, but you’re not wrong to call it out
If you do it early, you're at a disadvantage (unless you can market it heavily, which usually only is if you're the small player already).
If AWS unilaterally cut egress fees 100x, then a few things might happen:
GCP and Azure customers might migrate to AWS or have very serious phone calls with their account reps. (Hmm, GCP customers might be unable to reach an account rep.)
Third party providers, who are currently stuck in AWS, GCP and Azure because their customers can’t afford egress if they move out, will consider alternative hosting. This is IMO a big one, but it has nothing to do with whether GCP and Azure play along.
Some customers might shut down their colocated systems and move into AWS.
Why is being first a problem?
You just have to believe that, on the scale on AWS, 1Mbps in Europe costs 0.08$ (includes the transit, the hardware, the people etc). This includes every networking parts : do not forget the switches and stuff : VPC are free, and intra-zone network is free. Someone must pay, so it is sane to assume that "someone" is partly cross-region, partly internet egress, partly cross-zone.
Don't forget the R&D that revolves around VPC, subnets, security groups, acl etc.
So, yes, 0.08$/Mbps. Known from an external source which is a (partly) competitor. Trust me, dude. This number is legit.
There's a bunch of providers in Europe that will give you a whole server for €50/month with unmetered 1gbps networking. Or, as I like to consider it, €50/month for 1 gbps networking, and you get a free server.
Yeah, their networks may not have as much complexity and features as AWS, but I've never found any of that complexity appealing. Keep your network simple and let your hosts do the work.
It is the same for home internet connections : the offers is defined by the average use.
AWS and friends, on the other side, do not care about averages, and avoid you that burden too : what matter is how much you consumed, not "when" or "how" you consumed int
Per month?
Anyway, I can’t tell what your point is, but keep in mind that nearly all the complexity you discuss, and nearly all the hardware needed, applies to within-AZ traffic as well, and AWS doesn’t charge for that.
Also snarky statements such as the ones you are making do not particularly help your argument.
In many cases, this is way harder than it sounds.
You can also start the close account process which takes 60 days and just pay the fees for anything not torn down and then it all gets nuked at the end of that period.
I think interpreting this in a modern context, "religious" can be translated as "ethical"
I would say this is ever further from reality than perfect markets.
So we probably need both.
At least the US model often feels the need to produce something that appears sensible on the surface, although usually with gotchas in subsection 14 part C paragraph 2.
The EPA, NHTSA, FAA, FCC, SEC, etc go through periods of wax and wane with being undermined, then publicly embarrassed, then reconstructed with legitimate people, then undermined, etc. But they exist and ... eventually ... impose regulation.
I have never heard anyone say credibly as a politician to get rid of them, and goddamn are there a lot of corporate forces that want them gone, so as institutions these regulatory agencies have proven themselves.
The FDA however ... I'm pretty down on.
What is expensive is leaving in part. You can’t offload part of your data pipelines to, say, Digital Ocean or a colo cost-effectively.
There could probably be an automated tool for this that wouldn't be too complex.
[1] https://registry.terraform.io/providers/hashicorp/aws/latest...